ShopBack’s Justin Lee on why it let Meta’s Muse in while Amazon shut the door

Meta's Muse AI agent arrived in Singapore last month, connecting with ShopBack among other platforms. In contrast, Amazon barred Muse within a fortnight of its launch, citing concerns that an agent procuring purchases for a customer circumvented the personalized shopping experience Amazon intended. ShopBack, a cashback platform, opted to collaborate rather than exclude Muse.

Justin Lee, ShopBack's Chief Product Officer, explained that the web was designed for human reading, not for automated agents to act upon. In an interview with e27, Lee addressed ten questions regarding the implications of AI agents for a business that operates between shoppers and merchants. His responses reveal a confident stance, albeit with some unanswered questions.

ShopBack earns its revenue by providing rewards to shoppers, and Lee believed that agents offering rewards would keep them in the loop as the rewards are part of the best deal. Without a connector, Lee explained, the agent would shop like a person using a search engine, albeit more efficiently. The connector replaces this guesswork with direct access to ShopBack's live data, including merchants, prices, and associated savings.

Lee built the connector in just two weeks and described Muse as the next step in ShopBack's evolution, having previously transitioned from a website, to an app, to a browser extension, and finally into the ChatGPT App Store. Lee was asked whether Meta overlooked something Amazon had, but refrained from engaging in a direct confrontation.

He stated that different companies make different choices based on their business needs. Amazon controls the storefront, so it can afford to keep agents out. As a cashback platform, ShopBack doesn't own any shelves. If AI agents reshape retail as their backers anticipate, being the tool the agent uses is essential to prevent being bypassed.

Lee emphasized that the link and commission remain the same when Muse performs the clicking. The sale is counted as ShopBack's if it occurs through a ShopBack link, which Muse provides to the shopper. The merchant's tracking confirms the sale, and merchants pay based on confirmed sales under their existing terms. Lee clarified that even if an agent would have made a sale with another merchant, the reward from ShopBack tips the balance, making the reward the reason for the sale.

This aligns with the affiliate defense strategy, similar to Lazada's US$100 million affiliate initiative. Regarding customer ownership, Lee was clear about airlines, but vague about other entities. The traveler who books a flight through Muse purchases from the airline, which retains the relationship. Muse serves as the channel, while ShopBack acts as the rewards layer.

Meta and its Singapore partners have not addressed whether money changes hands with Meta and who retains conversational data. Lee explained that the agent weighs each deal's price and savings "on the facts." Users can verify the answer's accuracy by checking the ShopBack app, which provides the same deal. Although this offers a consistency check, it does not guarantee the cheapest deal across the entire market.

Lee addressed the pay-to-rank question, stating that merchants can already enhance rewards through ShopBack, and this boost translates into a greater saving for the shopper. This mirrors the promise that Muse does not consider ad budgets when determining deals. However, the line is thinner than it appears. Boosts represent merchant money spent to influence the agent's selection, working best when the agent prioritizes savings.

Brands concerned about AI agents failing to find them may find boosted cashback the most cost-effective means of being discovered. Lee clarified that the agent purchases from the merchant under their terms if the agent makes an incorrect booking. ShopBack handles only the rewards aspect, through its help center. This division of labor is both legally sound and a common practice in autonomous agent scenarios, although it exacerbates liability concerns.

Lee concluded that trust will ultimately determine whether people shop with agents. He emphasized the importance of clear distinctions between who does what in a market where Singaporean firms are leading AI agent deployment in APAC and also actively addressing agent rollbacks.